【Ken-AI-Part】关于基于AI的塔罗占测实例分享及可行性试探究

tarotk

先写个前言：
主要是这样，2个月前OpenAI发布了ChatGPT，一个AI的聊天机器人，他背后是OpenAI的GPT-3模型，起因具体看我朋友圈：
为了方便阅读都做成了手机屏幕大小的长屏截图
因为图片很多篇幅很长，重要的话都放大了字体方便跳过和跳转




附图如下：




跳转到这
然后想着神秘学领域呢，就拿着测了下塔罗，过程如下：
先是试着抽了日牌：




然后用最简单的圣三角牌阵尝试抽牌和解牌：


可以看到，结果非常的amazing，也太高accuracy哦不玄学了，这看着非常的合理啊，这让我对他的算法产生了质疑
关于算法部分我问了他一些问题，这是他的解答：（不想看的一直往下滑直接跳转）
让他抽牌，他说他不能物理意义上draw一张，666

I am sorry, as a language model AI, I don't have the ability to physically draw tarot cards. However, I can suggest a tarot card for you based on a random number generator. The card that has been selected for you is:

The Emperor

This card represents stability, authority, and structure. It suggests that you are in a place of power and control, and that you have the ability to make important decisions and take charge of your life. This card is associated with father figures, and may indicate that you are feeling a need for structure or guidance in your life. The Emperor can also suggest a time for taking action and making things happen, rather than just waiting for things to happen to you.

然后，关于他的随机数算法：

A random number generator is a mathematical tool used to generate a sequence of numbers that lack any pattern or predictability. The numbers generated by a random number generator are typically used in applications where a random outcome is desired, such as in statistical simulations, cryptography, and gaming.

There are various algorithms used to generate random numbers, ranging from simple linear congruential generators to more complex methods based on chaotic systems. The quality of the random numbers generated by a particular algorithm depends on the strength of the underlying mathematical method, as well as the quality of the implementation.

In some cases, random number generators may be seeded with an initial value, which is used to initialize the generator and to ensure that subsequent numbers generated are truly random. In other cases, random number generators may generate numbers based on a source of entropy, such as the current time or the movement of the mouse.

Regardless of the method used, the goal of a random number generator is to produce numbers that are truly random and that cannot be predicted or influenced by external factors. This makes them useful in a wide range of applications where a random outcome is desired.

他的算法是否具有统计学意义：

1. Yes, random number generators can have statistical significance. Random numbers are used in many statistical simulations and experiments, where they are used to model real-world processes or to generate data for analysis. For example, in Monte Carlo simulations, random numbers are used to generate sample data that can be used to estimate the behavior of a system or process.
   
2. 
   
3. In order for random numbers to have statistical significance, they must be generated using a high-quality random number generator that produces numbers that are truly random and that lack any pattern or predictability. This is important because patterns in the data generated by a random number generator can affect the results of statistical tests and simulations, leading to incorrect conclusions.
   
4. 
   
5. Additionally, the quality of the random numbers generated by a random number generator can be tested using statistical tests, such as the Chi-squared test or the Kolmogorov-Smirnov test, to ensure that they meet the required standards for statistical significance.
   
6. 
   
7. Overall, random number generators play a critical role in many areas of statistics and data analysis, and their quality and statistical significance are of utmost importance in ensuring the validity of the results obtained from statistical simulations and experiments.

复制代码

1. As an AI language model, I use a pseudo-random number generator to generate numbers. These numbers are not truly random, as they are based on a deterministic algorithm, but they do have some properties of randomness.
   
2. 
   
3. The pseudo-random numbers generated by my algorithm are designed to mimic the properties of truly random numbers, such as a lack of pattern or predictability. However, because they are based on a deterministic algorithm, they are not truly random and can be influenced by the initial seed value and other factors.
   
4. 
   
5. In some cases, the pseudo-random numbers generated by my algorithm may be suitable for certain applications where a random outcome is desired, such as in simple games or simulations. However, for applications where a truly random outcome is required, such as in cryptography or security, a truly random number generator should be used.

复制代码

跳转到这
我们看看那个结果，首先我确保两个图片也就是日牌和提问之间是不同的session，你们可以理解为确实是两个完全独立的过程，再鉴于他描述的算法随机性具有统计学意义的论证，我又再次看到塔罗的玄学哦不，应该是奇妙的联系性

虽然篇幅和时间关系我没有做更多的假设验证，直接给可行性结论是不可能也不科学的
但是这次尝试让我看到了一种新的可能和方向
别的不说，就单论实操和合理的解释性以及那奇妙的关联性，这可比很多半吊子占卜师强了好几个河蟹了lol

当然我们也看到了解牌的弱势，但可以说提供了一个很好的参考，而且最重要的，他给出了最终且直接的结论和建议
又有多少占卜师支支吾吾而不能决断
卜以夬疑，起码下定论和提供建议这块，再综合整体表现，我会给一个比较高的评价
最重要的是，他不是一个塔罗的模型，甚至本身没有关系，这只是非常顺带非常末微的功能，那如果是开发针对塔罗的AI呢？
喂塔罗的实战和数据呢？


未来可期啊

等等
谁说只能用于塔罗的?
我要去其他区玩啦
然后还是给东方稍微留一个净土吧


未完待续...
//也不一定会续...我留的坑没填的还少吗XD
@Delilah


原则上讲在AI的帮助下塔罗师的占卜能力和水平应该会逐步提高甚至飞跃
但是悲观的来看
未来会出现更糟糕的, 更多招摇撞骗的网络占卜
甚至完全小白, 甚至直接说是AI占卜的噱头

人工智能
会有命运吗?



补充内容 (2023-2-10 03:41):
严格意义上讲，很多人没看仔细，这个不是ChatGPT
我的demo是基于davinci构建的类似原生ChatGPT
davinci-003和chatgpt都是从davinci-002改进，但却截然不同

补充内容 (2023-2-10 03:42):
ChatGPT牺牲了几乎所有的上下文能力来换取与人类对齐，更像个人类在对话
davinci也就是我这个版本保留了强大的调整和学习能力，提高了零样本的效能

补充内容 (2023-2-10 03:44):
所以ChatGPT会拒绝知识外的问题并且反复强调他是个AI
而我的demo可以通过输入来进一步调整和学习

所以很可能用ChatGPT复现不了我的结果或者偏差较大

Lenny

哈哈哈，有趣的尝试，确实比某些连牌意都背不明白的小白占卜师厉害了。不知道它还能不能做更多牌的牌阵来占卜，比如凯尔特十字，需要结合2~3个位置的牌一起解读的整合能力。

尼玛利亚瑟王

《从AI算命到给AI算命》

奥村英二

AI扶乩在这个之前就有了

有为有不为

用这个AI可以开一个起名算卦的店了

magical656

wow！！！确实比好多新手还精准直接！

tarotk

Lenny 发表于 2023-2-9 10:02
哈哈哈，有趣的尝试，确实比某些连牌意都背不明白的小白占卜师厉害了。不知道它还能不能做更多牌的牌阵来占 ...

比一些只能记住牌但缺无法推理和判断的也强太多了

基本的推理反而是强项

tarotk

Lenny 发表于 2023-2-9 10:02
哈哈哈，有趣的尝试，确实比某些连牌意都背不明白的小白占卜师厉害了。不知道它还能不能做更多牌的牌阵来占 ...

只要你定义好牌阵，并告诉他牌阵各个位置的关联和意义
他可以学习一种全新的甚至不存在的牌阵
并进行推演

tarotk

尼玛利亚瑟王 发表于 2023-2-9 13:03
《从AI算命到给AI算命》

tarotk

奥村英二 发表于 2023-2-9 13:35
AI扶乩在这个之前就有了

但根本不是一回事